On June 30, 2026 (local time), the U.S. Department of Commerce formally lifted export controls on Anthropic's AI models Claude Fable 5 and Mythos 5, ending a roughly 18-day global access suspension. The company said it would begin restoring access in stages from July 1.
June 30, 2026 · US Dept. of Commerce · Anthropic
US Lifts Export Ban on Anthropic's Top-Tier Claude Models After 18-Day Freeze
Commerce withdrew the export controls on Claude Fable 5 and Mythos 5 that had halted global access since June 12 over cyberwar misuse fears. Fable 5 begins restoring July 1; Mythos 5 returns only for vetted US institutions.
18
days of global suspension (Jun 12 → Jun 30)
1M
token context window on both models
1st
"deemed export" ever applied to a commercial AI model
Timeline of the suspension
JUN 9
Fable 5 & Mythos 5 released — Mythos-class, above Opus 4.8
→
JUN 12 · 5:21 PM ET
Export order bars foreign-national access over cyberwar misuse fears
→
JUN 30
Commerce withdraws controls — no license required; Fable 5 restores Jul 1
Fable 5 — public
Safety classifiers detect & block harmful queries
Auto fallback to Opus 4.8 on detected risk
Restoring globally from July 1
Praised for reasoning & long-horizon agentic work
Mythos 5 — restricted
Relaxed classifiers in cyber & biology domains
Among strongest cybersecurity capabilities in the world
Delivered via Project Glasswing to vetted partners
Returns only for vetted US institutions
The unresolved tension
The ban was driven by fears Fable 5's guardrails could be jailbroken for zero-day discovery and exploit construction.
Backdrop: Warnings from Amazon CEO Andy Jassy and an NSA review.
Still contested: Stricter cyber filters block most bypasses — but developers warn of over-blocking legitimate debugging.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…